Try it
See how your box would answer.
Pick a model and chat. Replies come from a real model we self-host, and they stream at the real tokens per second Base Router delivers for that model, so the feel is honest.
Box
Base RouterJetson Orin Nano Super · GPUModel
tok/s = tokens per second, roughly how fast words appear · 8B = model size in billions of parameters (bigger is usually smarter, but slower)
generation speed
38.0tok/s
≈ how fast replies appear
Liquid LFM2.5 1.2B on Base Router
offlinebase.locallocal · offline
Pick a model, then ask something. The reply streams at the box’s real tok/s rate.